Bengaluru-based AI startup OnFinance has launched a language model named NeoGPT to tailor AI solutions for the banking, financial services, and insurance (BFSI) sectors. The company aims to leverage AI to bring transformative change to financial services, focusing on security and efficiency. OnFinance is the first generative AI startup to onboard the National Stock Exchange of India (NSE) as a client. The applications of NeoGPT include financial research, equity research, underwriting, compliance, and wealth advisory. The company ensures data security by using on-prem solutions and not using customer data for training or analysis. OnFinance builds its models on robust open-source frameworks and uses Q&A datasets derived from financial news sources for instruction tuning.
Since the release of Anthropic’s Claude 3.5 model family, social media platforms, particularly X, have been all about Claude 3.5 Sonnet. It enhances user interaction by providing a dedicated window alongside conversations and significantly improves data interpretation and visualization capabilities. Swami Sivasubramanian, head of AI services and data at AWS, highlighted Claude 3.5 Sonnet's strengths for data science and analysis, as well as vision capabilities. Users on X have praised GPT 4o's data visualization capabilities, but there were limitations mentioned as well. Both Claude 3.5 Sonnet and GPT-4o showed capabilities in data analysis, with strengths and weaknesses for each. A research paper by Generative AI Research Lab showed that GPT-4o slightly outperforms Claude-3.5-Sonnet in overall visual reasoning tasks, but the difference is minimal. There is no established benchmark for data visualization evaluation, but some research papers have developed benchmarks for data evaluation in LLMs. LLMs struggle with complex visualizations requiring multiple visual channels, and performance decreases with increasing query complexity. The article provides examples and user experiences to support the claims.
ChatGPT 4.0, the latest version of the Chatbot developed by Open AI, has been released. It can now analyze images and provide descriptions, making it useful for visually impaired individuals. ChatGPT 4.0 can also provide emotional support and engage in meaningful conversations, offering companionship to those who feel isolated. However, over-reliance on AI for social interaction may lead to the deterioration of human relationships and emotional isolation. Users should be conscious of the time spent interacting with AI and ensure it does not replace genuine human connections. The development of AI technology raises questions about transparency, information sources, and the establishment of legal guardrails. Despite the risks, ChatGPT 4.0 is considered a significant advancement in technology.
OpenAI recently announced the release of a prototype AI tool called SearchGPT, which aims to provide web search capabilities and answer questions. However, in a demo, the tool provided incorrect information about music festivals in Boone, North Carolina in August. OpenAI acknowledged the error and stated that they will continue to improve the tool. The long-term goal is to incorporate search features into ChatGPT, OpenAI's flagship AI product. While the festival mix-up is seen as a minor issue, it highlights the challenges of AI-powered search. Previous incidents of AI self-owns in the field of generative AI, including racist responses from ChatGPT and errors from Google Bard, have occurred. AI-powered search engines often produce flawed answers due to the nature of prediction-based models. Concerns also arise about the impact on website traffic and revenue as AI tools summarize information from original sources. OpenAI has partnered with media publishers to drive traffic to external sites. The cycle of tech companies releasing flawed products and promising improvements continues in the AI boom. [9ca06533]